Complements and Meat Demand in the U.S. Christopher Davis1, Stela Stefanova2, William Hahn3, and Steven Yen4

1Economist with the Markets and Trade Economics Division, Economic Research Service, U.S. Department of Agriculture. 1800 M Street, NW, Washington D.C. 20036. Phone:

(202) 694-5167. Email Address: chrisdavis@ers.usda.gov 2Graduate Assistant in the Department of Economics,

the University of Delaware, Newark, Delaware. Phone: (302) 831-6424. Email address: stefstel@UDel.Edu

3Senior Economist with the Markets and Trade Economics Division, Economic Research Service, U.S. Department of Agriculture. 1800 M Street, NW, Washington D.C. 20036. Phone:

(202) 694-5175. Email Address: whahn@ers.usda.gov 4Associate Professor in the Department of Agricultural Economics,

the University of Tennessee, Knoxville, Tennessee. Phone: (865) 974-7474. Email address: syen@utk.edu

Selected Paper prepared for presentation at the American Agricultural Economics Association Annual Meeting, Orlando, Florida, July 27- 29, 2008.

1

Abstract

In this study we estimated the price elasticities among meats, vegetables, grains, and potatoes and the impact that different levels of income have on the demand for these commodities. The 2005 Nielsen retail home scan data were used to construct a censored demand system of 14 equations. Results revealed that the uncompensated cross-price elasticities for both low and high-incomes suggest both substitution and complement relationships, while the compensated price elasticities are dominated primarily by substitution relationships. Our findings also revealed that expenditure elasticities among both low and high-income households differ for most commodities1.

Key Words: censored dependent variables, meats, poultry, fish, vegetables, sample selection

model, two-step estimation.

JEL Classifications: C25, D12, Q11

1 Disclaimer: The opinions and analysis presented represent the authors’ idea and do not necessarily reflect Economic Research Service or the U.S. Department of Agriculture position.

Copyright 2008 by Christopher Davis, Stela Stefanova, William Hahn, and Steven Yen. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies.

2

2

Complements and Meat Demand in the U.S.

Introduction

For decades consumer food demand studies have provided insightful information on who eats

what, where, and how much (Yen, Lin and Davis, 2008; Davis and Lin, 2005a, 2005b; Lin et al.,

2003). Many studies have addressed the substitutability of meats like beef, pork, poultry, and

fish by types, by parts and/or cuts, as well as assessed complementary relationships among them

(Thompson, 2004; Kinnucan et al., 1997; Eales and Unnevehr, 1993; Capps and Lambregts,

1991; Moschini and Meilke, 1989; Cheng and Capps, 1988; Purcell and Raunikar, 1971).

Likewise, there have been several studies conducted that analyze fruit, vegetable, and grain

demands in the United States through the estimation of price elasticities and income or

expenditure elasticity (Just and Weninger, 1997). Separately these analyses have helped food

companies improve their supply-chain management and enhanced distributors’ marketing

strategies to increase the flow of goods.

Several recent studies have emphasized the substitution and complementarity relationships

between meats and other dishes (Thompson, 2004; Kinnucan et al., 1997; Park et al., 1996). For

most non-vegetarian households, the demands for vegetables, grains, and potatoes are contingent

upon the demands for beef, pork, poultry, or fish. Theoretically, complementary relationships

between meats, vegetables, grains, and potatoes are essential components that help identify a

consumer’s food consumption basket.

3

3

Park et al. (1996) examined 12 food commodity groups (food away from home, beef, pork,

chicken, fish, cheese, milk, fruits, vegetables, breakfast cereals, bread, and fats and oils) to

determine whether commodity demand projections are based on individual income strata rather

than on average estimates of price and income elasticities. Results from a 1987-88 Nationwide

Food Consumption Survey reveal that own-price elasticities are similar between income groups

for most commodities and income elasticities are consistently higher for the lower-income group.

Like Park et al. (1996), we analyzed the impact that different levels of income have on

commodity demand, but also estimated the price elasticities among meats, vegetables, grains,

and potatoes using the 2005 Nielsen retail home scan data. For this study special attention is

given to cross-price relationship, particularly complementarities among the commodity groups.

A censored demand system of 14 equations is employed to determine the price and

income/expenditure relationships that exist among a selected group of commodities.

Methodology We estimated a large censored demand system using a multivariate sample selection model

developed by Yen and Lin (2006), which was estimated with a two step procedure proposed by

Shonkwiler and Yen (1999). This approach accommodates zero purchases and simplifies the

computational burden, while still producing consistent estimates. We followed closely the

specification of Yen, Lin and Davis (2008).

We assumed that meat products and hypothesized complementary products are separable from

all other goods. In the first step of the procedure, censoring of each commodity is governed by

the following stochastic sample selection process.

4

4

[ ]( ; ) , 1,...,i i i iw d f x e i nθ= + = , (1)

where id = 1 if ' 0i iz uγ + > and id = 0 if ' 0i iz uγ + ≤ , wi is the expenditure share of the ith

commodity, x and z are vectors of explanatory variables, θ and γi are vectors of parameters, and

ei and ui are random errors.

Assuming the translog utility function, the translog demand system in expenditure form can be

derived as:

1

1 1 1

log, 1,...,

log

n

i jijj

i n n n

j jkjj k j

vw i n

v

α β

α β

=

= = =

+= =

+

∑

∑ ∑∑, (2)

where jv are expenditure normalized prices for commodity j. Homogeneity is implied in the

above equation by the use of the normalized prices for all commodities, and symmetry is

imposed with the restrictions

,ij ji i jβ β= ∀ (3)

We allowed the intercept iα to vary with demographic variables hl such that

01

, 1,... 1L

i i i h i nα α α=

= + = −∑ l l

l

(4)

One issue with the censored system approach specified above is that the adding-up restriction

cannot be imposed. Following the approach suggested by Yen and Lin (2006), we estimated the

first n –1 equations and calculate elasticities for the nth equation using the adding-up property in

demand theory. Even though the estimates are not invariant to the equation excluded, Yen and

Lin (2006) showed that the elasticity estimates are stable regardless of which commodity is

treated as the residual category.

5

5

The system of demand equations in share form can be written as:

( ) ( ) ( ; ) ( )i i i i i i i i i iw E w z f x zξ γ θ δ φ γ ξ′ ′= + = Φ + + , (5)

where iδ is the covariance between the error terms ei and ui, ( )i iz γ′Φ and ( )i iz γ′Φ are the normal

cumulative distribution and probability density functions respectively, and ( )i ii w E wξ = − is a

heteroskedastic error term, with ( ) 0iE ξ = (Shonkwiler and Yen, 1999). The system can be

estimated using the two step procedure. First, we obtained maximum-likelihood (ML) estimates

for iγ based on binary probit for iw = 0 and iw > 0. Second, assuming that the disturbances

( ie , iu ) are distributed bivariate normal with cov( ie , iu ) = iδ , we estimated the demand

parameters θ and covariances iδ in the system

ˆ ˆ( ) ( ; ) ( )i ii i i i i iw z f x zγ θ δ φ γ ξ′ ′= Φ + + (6)

using iterated seemingly unrelated regressions. Demand elasticities for the n–1 goods can be

derived by differentiating Equation (6). Elasticities for the residual good are calculated using the

adding up restriction. To derive compensated demand elasticites, we used Slutsky’s equation.

Data

The ACNielsen 2005 Homescan data contain demographic and food purchases information for a

nationwide panel of representative households. Each household in the panel is given a hand-held

scanning device to scan at home all food items purchased at any retail outlet. Some households

record only UPC coded foods while others scan both UPC coded as well as random weight

items. In this study we used the smaller subset of 8,216 households that recorded both UPC-

coded and random weight products. These households reported 7,597,426 purchases in 2005:

6

6

4,001,639 dry grocery products; 1,379,832 random weight items; 900,100 dairy products and

1,315,855 produce, meat and frozen food. Each purchase record contains data on product

characteristics, quantity purchased, price paid with and without promotions, date of purchase,

store and brand information and is matched to a household record. Information on size and

composition of household, income, origin, age, race, gender, education and occupation of

household members, as well as market location data is available for each household. Projection

factors (sample weights) are provided by ACNielsen to be used at the household level to provide

representative estimates for the U.S. population.

It is not feasible to simultaneously account for the joint consumption of all possible types of

food, so we further limited the dataset to only foods that people are more likely to consume

together with meat or fish. Using What We Eat in America (WWEIA)-NHANES data for years

1999-2000, 2001-2002, and 2003-2004, we compared the probabilities of consuming different

types of food conditional on consuming different types of meat and fish with the marginal

probabilities of consuming every food category. For our analysis, we aggregated these food

categories by the first two digits of the NHANES food codes. Table 1 provides the first two

digits of the food codes and the conditional and unconditional probabilities calculated from

2003-2004 data. The results are very similar for the other periods. We identified 9 food

categories - cheese, bread, rice, pasta products, tomato products, orange vegetables, potatoes,

salad and hamburger vegetables, and all other vegetables, that we expected to be complementary

to meat and fish.

7

7

The final dataset used in this study is compiled from ACNielsen Homescan data and includes all

purchase records of beef, pork, poultry, fish, cheese, bread, rice, pasta products, tomato products,

and vegetables made by consumers reporting random weight items. Prices are measured for all

products in dollars per pound after any coupons or promotion information is taken into account.

Meat products, mostly fresh meats, which could not be readily identified as specific meat types,

are assigned into groups using the descriptions of the UPC and designated codes for the random

weight items. Meat is categorized into five groups: beef, pork, poultry, fish, and processed meats.

The Processed meats include sausages, corn dogs, hotdogs, salami, mixed meats, lunch meats,

bratwursts, refrigerated bacon, sandwich steak, canned meat and deviled ham. Beef, pork,

poultry, and fish are broad categories including products in all forms -- fresh, canned and frozen

-- that could be identified in the respective categories.

Similar to Yen, Lin, and Davis (2008), we split the sample in low and high income groups using

a cutoff of 350% poverty income ratio (PIR). The PIR is calculated by dividing the mid-point of

the range of self reported income to the Federal poverty thresholds, i.e. it expresses income as a

percent of the Federal poverty thresholds for households of different size. The 350% cutoff is

used to generate similar number of observations in the two groups. The sample sizes of the low

income and the high income groups are 4,129 and 4,087 respectively.

We focused on 14 types of food products. To organize the food categories, we first identified the

product modules that fall within each category. We converted the quantities purchased to pounds

and prices to dollars. Purchased quantities and expenditures on each product category are

aggregated to the annual level for each household. The price for each product category is

8

8

calculated by dividing the expenditures by the purchase quantity, thereby representing unit

values. Missing prices for non-consuming households are assigned mean unit values calculated

by region.

Descriptive statistics for the 14 chosen food categories are provided in Table 2 and Table 3.

Almost all of the sample households consume some type of bread. Cheese, processed meats and

vegetables are the next most popular food categories with over 95% of the households

consuming. This is followed by the potato category, which is consumed by 93% of the

households. Beef, fish, poultry, noodles and orange vegetables are consumed by about 90% of

the households. About 87% purchase pork and some type of a tomato product. Rice is purchased

by the least number of households – about 75%. The consumption patterns are not dramatically

different between the low and the high income households. However, quantities consumed are

somewhat different between the two income samples. Notable differences are that the high

income households consuming more fish, orange and salad vegetables and less beef, pork,

processed meats, bread and potatoes when compared to the low income households.

The prices paid by the high income households are higher than those paid by the low income

households for all product categories. Among the 14 food products, fish and cheese are the most

expensive and tomato products the least expensive per pound. Poultry meat is the least expensive

meat product.

Table 4 presents descriptive statistics for certain demographic variables. Following Yen, Lin, and

Davis (2008), household size, dummy variables for age groups, race and regional dummies,

9

9

presence of children, and dummy variables for marital status correspond to a set of explanatory

variables used in the second stage regressions. All these demographic variables plus female head

of household employment, male head of household employment, dummy variables for levels of

education, and the poverty income ratio are included in the first step probit regressions. The

average household size for the low income sample is 2.50 and is 2.22 for the high income

sample. More than 50% of the households in both income categories are headed by individuals

between 40 and 60 years of age. Geographic characteristics of the respondents are represented by

regional variables. The low and high income distributions are similar with slightly more low

income households located in the central and south regions of the U.S. The racial and ethnic

characteristics are also similar across income groups with 77% White, 14% Black, 2% Asian, 9%

Hispanic, and 7% other race for the low income sample. The low income sample includes higher

percent ages of male and female head of household employed – 60% and 56% respectively,

compared to the high income group - 40% and 38%. Individuals in the high income group are

more educated with 55% of the sample having college education and above, compared with only

27% of the low income sample in the same category.

Demand Elasticities High-Income Households

Results of the uncompensated price and expenditure elasticities for high-income households are

found in table 5. All uncompensated own-price elasticities are negative and statistically

significant at the 1 percent level. Own-price elasticities for meat products ranged from –1.22 for

beef to –1.47 for poultry, while own-price elasticities for side dishes ranged from –1.19 for

tomatoes to –2.23 for salad vegetables.

10

10

Seventy-three percent (133 out of 182) of the uncompensated cross-price elasticities are

significant at the 10 percent level or lower, which suggest a mixture of gross complements and

substitutes among the meat product categories. With an exception of processed meats and beef,

and poultry and pork relationships, all other meats (including poultry and fish) combinations are

gross substitutes for each other. In identifying meat products that serve as complements to side

dishes, our results reveal that beef & noodles are negatively related, which means a 1 percent

increase in the price of beef will result in a .05 percent decrease in the demand for noodles. Other

complementary relations include pork & tomatoes, pork & salads, fish & orange vegetables, and

processed meat & other vegetables.

The expenditure elasticities vary slightly, ranging from 0.92 for cheese, to 1.03 for pork, rice,

and potatoes (table 6). Expenditure elasticities for meats, particularly beef, pork, and poultry are

above unity, which are similar to the aggregate meat expenditure elaticities derived by Kesavan

et al. (1993). For side dishes such as rice, potatoes, and other vegetables, expenditure elasticities

also exceeded one which differs from Parks et al. (1996) findings for aggregated vegetables for

poverty and non-poverty status households. All expenditure elasticities are estimated with

relatively high precision (with small standard errors), relative to the cross-price elasticities.

Table 7 presents the compensated price elasticities for high-income households. Similar to their

Marshallian counterparts, all of the compensated own-price elasticities are significant, with the

appropriate signs put forward by economic theory. All compensated own-price elasticities are

greater than unity, ranging from –1.16 for beef to –1.37 for poultry, and from –1.13 for tomatoes

11

11

to –2.17 for rice. The own-price elasticity of pork (–1.20), beef (–1.16), fish (–1.23), and poultry

(–1.37) are all elastic and more price responsive than the Park et al. (1996) own-price elasticity

estimates for the same commodities for non-poverty status households. Contrary to

uncompensated elasticities, which suggest gross substitutes and complements among meat

products and side dishes, the compensated elaticities suggest predominantly gross substitutes,

minus a couple exceptions. A strong substitution relationship between beef and potatoes, pork

and rice, poultry and orange vegetables, and fish and salad vegetables indicates that vegetarian

diets can be easily exchanged for diets that include meat for some high-income households.

Low-Income Households

The uncompensated price and total expenditure elasticities for low-income households are

presented in table 8. All uncompensated own-price elasticities are significant at the 1 percent

level with negative signs similar to elasticities for the high-income households. Of the 14

commodities analyzed, all are greater than unity, ranging from –1.32 for fish to –1.74 for

processed meat, to –2.43 for rice. Unlike Park et al. (1996), our own-price elasticities for beef,

pork, fish, vegetables, and bread are elastic for low-income households or poverty households

but more responsive to changes in price.

Of the 184 possible cross-price elasticities, 121 (60%) are positive and statistically significant.

This is similar to Nayga and Caps’ (1994) study in which a majority of the cross-price

elasiticities had positive signs implying a substitution relationship. The total expenditure

elasticities for the 14 commodities are precisely estimated as in the low-income sample (table 9).

All expenditure elasticities are statistically significant at the 1 percent level with positive signs

12

12

implying a direct relationship between commodity consumption and increases in expenditures.

Commodities such as beef, pork, fish, poultry, rice, tomatoes orange vegetables, and potatoes are

all above unity, which differs from Park et al. (1996), indicating a greater level of sensitivity to

changes in consumers’ expenditures. In contrast, the expenditure elasticity for bread is

considerably less than unity and is similar to what Park et al. (1996) found for poverty

households. Along with bread, there are other commodities (processed meats, cheese, noodles,

other vegetables and salad vegetables) with expenditure elasticities less than unity.

Table 10 presents the compensated price elasticities for low-income households. All

compensated own-price elaticities are significant, negative and greater than unity, ranging from

–1.28 for fish to –1.70 for processed meat, to –2.40 for rice. The large own-price elasticity for

rice implies that it is highly sensitive to changes in price. Like results from the high-income

household sample, the low-income household sample cross-price elasticities suggest s net

substitution relationships in most cases, 174 of which are significant and positive. The one

significant complementary relationship is among fish and salad vegetables, where a 1 percent

increase in the price of fish will only reduce the demand for salad vegetable marginally (.04

percent).

Conclusion and Recommendations

In this study we analyzed at-home consumption of different types for meat products and side

dishes such as rice, vegetables, bread, and cheese by estimating a 14 equation demand system

using a multivariate sample selection model. Nielsen household Homescan data highlighted zero

purchases of many commodities due to some disaggregate commodities and commodities that

13

13

were not purchased during specific time periods. We estimated this large censored demand

system using the Heckman two-step procedure. The two-step procedure produces statistically

consistent estimates although it is less efficient than the ML alternative of Yen and Lin (2006).

This study presented information on the demand for beef, pork, fish, poultry, vegetables, cheese,

bread, and rice separately for low-income and high-income households. Findings indicated

notable differences in the elasticity estimates between the two groups of households, particularly

with compensated and uncompensated price elasticities. Our chief objective for this study was

based on a perceived logic guiding consumers’ purchase behaviors. That logic is that non-

vegetarians seek to purchase a desired meat (beef, pork, fish, or poultry) first and then choose a

side dish to accompany their choice of meat or vice versa. Results revealed that the

uncompensated cross-price elasticities for both low and high incomes suggest both substitution

and complement relationships, while the compensated price elasticities are dominated primarily

by substitution relationships. These findings are surprising and unexpected given hypothesized

logic consumers often use when planning meals or shopping for food for family members. This

perceived logic is rejected by the results found for compensated price elasticities, but not rejected

based on the results rendered from the uncompensated price elasticities.

Elasticities derived from a massive database like the Nielsen retail homescan data can prove to

be quite useful. Agricultural industries or food retailers may use information from studies such

as this to boost sales through advertisement of specific food combinations, especially among

frozen meals. Elasticities from this study can be used to determine the impact increased demand

for corn and soybeans used in other agricultural production will have on consumer demand for

14

14

specific meats primarily pork, poultry, and fish. We anticipate that as input or feed prices rise in

response to increasing demand for corn and/or soybeans for other agricultural production,

demand elasticities derived from this study for meats can serve as a useful tool in addressing

changes in consumer demand for food commodities and making policy recommendations.

15

15

References

Capps, O. Jr. and J.A. Lambregts. “Assessing Effects of Prices and Advertising on Purchases of Finfish and Shellfish in a Local Market in Texas.” Southern Journal of Agricultural Economics. 23(1) (July 1991): 181-194.

Cheng, H. and O. Capps, Jr. “Demand Analysis of Fresh and Frozen Finfish and Shellfish

in the United States.” American Journal of Agricultural Economics. 70(3) (August 1988): 533-542.

Davis, C.G., Lin, B., 2005a. Factors affecting U.S. beef consumption. Washington DC: US

Department of Agriculture, Economic Research Service, Electronic Outlook Report from the Economic Research Service, LDP-M-135-02, October.

Davis, C.G., Lin, B., 2005b. Factors affecting U.S. pork consumption. Washington DC: US

Department of Agriculture, Economic Research Service, Electronic Outlook Report from the Economic Research Service, LDP-M-130-01, May.

Eales, J.S. and L.J. Unnevehr. “Simultaneity and Structural Change in U.S. Meat

Demand.” American Journal of Agricultural Economics. 75(2) (May 1993): 259 -268.

Just, R.E, and W. Weninger. “Economic evaluation of the farmers' market nutrition

program.” American Journal of Agricultural Economics. 79, no. 3 (Aug 1997): p. 902-917.

Kesavan, T., Z. A. Hassan, H. H. Jensen, and S.R. Johnson. 1993. “Dynamics and Long-run Structure in U.S. Meat Demand.” Canadian Journal of Agricultural Economics 41(2): 139–153.

Kinnucan, H.W., H. Xiao, C. Hsia, and J.D. Jackson. “Effects of Health Information and

Generic Advertising on U.S. Meat Demand.” American Journal of Agricultural Economics. 79(1) (February 1997): 13-23.

Lin B.H., Variyam, J., Allshouse, J., Cromartie, J., 2003. Food and Agricultural Commodity Consumption in the United States: Looking ahead to 2020. Washington DC: US Department of Agriculture, Agricultural Economic Report No. 820, February. Available at: http://www.ers.usda.gov/publications/aer820/ aer820.pdf. Accessed 21 October 2005.

Moschini, G. and K. D. Meilke. “Modeling the Pattern of Structural Change in U.S. Meat

Demand.” American Journal of Agricultural Economics. 71(2) (May 1989): 253-261. Nayga, R.M, and O. Capps, Jr. 1994. “Tests of Weak Separability in Disaggregated Meat

Products.” American Journal of Agricultural Economics 76(4):800–08.

16

16

Park, J.L., R.B. Holcomb, K.C. Raper, and O.Capps, Jr. 1996. “A Demand Systems Analysis of

Food Commodities by U.S. Households Segmented by Income.” American Journal of Agricultural Economics 78(2): 290–300.

Purcell, C.P. and R. Raunikar. “Price Elasticities from Panel Data: Meat, Poultry, and

Fish.” American Journal of Agricultural Economics. 53(2) (May 1971): 216- 221.

Shonkwiler , J.S. and S. T. Yen. 1999. “Two-Step Estimation of a Censored System of

Equations.” American Journal of Agricultural Economics 81(4): 972-982. Thompson, W. 2004. “Using Elasticities From An Almost Ideal Demand System? Watch Out

For Group Expenditure.” American Journal of Agricultural Economics 86(4): 1108–1116.

Yen, S.T., B. Lin, and C. Davis. “Consumer Knowledge and Meat Consumption at Home and Away From Home.” Food Policy. In press 2008.

Yen, S.T., and B. Lin. 2006. “A Sample Selection Approach to Censored Demand Systems.” American Journal of Agricultural Economics 88(3): 742-749.

17

17

Table1: Conditional vs. Unconditional Probabilities of Consumption Unconditional

Probability Conditional probability

Food Type 2-digit code P(Type)

P(Type/ Fish)

P(Type/ Beef)

P(Type/ Poultry)

P(Type/ Pork)

Cheese 14 3.19 10.74 39.91 11.70 15.22 Breads 51 5.86 24.91 52.89 23.71 35.08 Biscuits 52 1.80 10.17 20.98 14.50 14.77 Pasta products and rice 56 1.35 15.09 7.15 10.32 10.27 Potatoes 71 3.65 27.89 35.86 32.90 21.96 Raw green vegetables 72 0.59 12.00 4.99 7.73 6.00 Orange vegetables 73 0.70 7.89 6.94 8.47 5.47 Tomato products 74 4.13 49.83 66.93 37.61 28.34 Hamburger vegetables 75 6.38 75.66 95.27 58.49 51.87

18

18

Table2: Average Annual Household Purchases: Prices, Expenditures and Quantity by Income Low Income High Income Variable Mean Std.Dev. Mean Std.Dev.Prices Beef 3.06 1.09 3.69 1.47Pork 3.40 2.55 3.88 2.89Fish 3.81 1.95 4.74 2.33Poultry 2.05 0.93 2.32 1.06Processed Meat 3.57 1.35 4.27 1.64Cheese 3.92 1.18 4.56 1.50Bread 1.54 0.76 1.83 0.99Rice 1.92 0.95 2.12 1.07Noodles 1.23 0.80 1.43 0.83Tomato Products 0.79 0.32 0.88 0.61Orange 1.02 0.39 1.13 0.49Potatoes 1.06 0.71 1.20 0.78Other Vegetables 1.89 1.90 2.36 1.98Salad Vegetables 1.28 0.51 1.48 0.60Expenditures Beef 119.79 133.40 134.67 156.52Pork 68.74 76.53 70.56 77.67Fish 48.96 70.22 68.30 101.47Poultry 66.03 72.89 71.15 77.29Processed Meat 103.36 101.71 105.46 109.06Cheese 88.62 80.86 105.21 96.74Bread 90.79 70.37 94.34 75.31Rice 8.05 12.62 8.21 11.39Noodles 15.00 18.96 15.77 20.10Tomato Products 11.63 14.50 11.12 13.02Orange 13.85 17.02 16.69 22.46Potatoes 27.72 27.84 25.08 26.30Other Vegetables 59.79 52.79 73.71 67.05Salad Vegetables 50.64 53.50 63.26 65.35Quantities Beef 44.32 48.63 40.40 44.94Pork 30.56 38.40 27.29 33.82Fish 13.78 18.22 15.48 21.05Poultry 45.26 51.98 42.42 49.37Processed Meat 35.03 33.63 29.69 29.12Cheese 25.59 23.51 26.81 24.21Bread 83.20 92.47 74.99 76.18Rice 8.03 19.69 6.93 14.94Noodles 15.26 20.07 13.60 21.05Tomato Products 17.08 22.23 15.01 18.59

19

19

Orange 15.94 19.83 17.49 26.23Potatoes 49.91 51.89 37.83 40.51Other Vegetables 55.03 49.36 54.61 47.74Salad Vegetables 49.86 52.97 53.67 53.86

20

20

Table 3: Consuming Households

Food Type

Number Households Consuming

Total %

consuming Low Income % consuming

High Income %

Consuming Beef 7450 90.68 90.87 90.48 Pork 7164 87.2 87.82 86.44 Fish 7451 90.69 90.82 90.36 Poultry 7420 90.31 90.14 90.41 Processed Meat 7947 96.73 97.00 96.43 Cheese 8048 97.96 97.80 98.12 Bread 8145 99.14 99.25 99.00 Rice 6198 75.44 75.15 75.43 Noodles 7344 89.39 89.25 89.33 Tomato Products 7178 87.37 88.21 86.35 Orange Vegetables 7383 89.86 89.42 90.26 Potatoes 7679 93.46 94.14 92.76 Other Vegetables 8061 98.11 98.06 98.14 Salad Vegetables 7950 96.76 96.83 96.70

21

21

Table 4: Sample Statistics for Demographic Variables Variable Low High incomeHousehold Size 2.50 (1.50)* 2.22 (1.11)*Presence of children 0.29 0.18Age of household head < 40 0.15 0.15Age of household head 40-60 0.57 0.70Age of household head > 60 0.28 0.14Married 0.53 0.61Widowed 0.13 0.06Divorced 0.18 0.13Single 0.17 0.20East 0.22 0.23Central 0.18 0.15South 0.40 0.37West 0.20 0.25Education < high school 0.05 0.01High School 0.31 0.15Some College 0.36 0.29College and above 0.27 0.55White 0.77 0.76Black 0.14 0.13Asian 0.02 0.05Hispanic 0.09 0.07Other race 0.07 0.06Male head employed 0.60 0.40Female head employed 0.56 0.38Poverty income ratio 212 (81)* 623 (266)*Sample size 4129 4087* Standard errors in parentheses

22

22

Table 5: Uncompensated Price Elasticites: High Income Households Variable Beef Pork Fish Poultry Processed Cheese Bread Rice Noodles Tomatoes OrangeV Potatoes OtherV SaladV Beef -1.22*** -0.01 0.03*** 0.03** -0.01** 0.00 -0.01 0.05*** -0.02*** 0.05*** 0.01 0.04*** -0.01 0.03*** Pork -0.02 -1.29*** 0.00 -0.06*** 0.04*** 0.04*** 0.11*** 0.07*** 0.00 -0.05** 0.03** 0.07*** 0.03** -0.01 Fish 0.08*** 0.00 -1.27*** 0.04* 0.04*** 0.00 0.00 0.05*** 0.02 -0.03 -0.03* 0.04*** 0.05*** 0.01 Poultry 0.03** -0.03*** 0.02* -1.47*** 0.02*** 0.01* 0.07*** 0.05*** 0.02** 0.07*** 0.04*** 0.06*** 0.04*** 0.02* Processed -0.05* 0.08*** 0.06*** 0.08*** -1.40*** -0.04 0.06** 0.13*** 0.00 0.06** 0.04* 0.05*** -0.03** 0.00 Cheese 0.01 0.08*** 0.01 0.06** -0.03 -1.57*** 0.13*** 0.16*** 0.03 -0.02 0.02 0.10*** -0.04*** 0.16*** Bread -0.01* 0.10*** 0.01 0.13*** 0.03*** 0.06*** -1.52*** 0.08*** 0.05*** -0.01 0.06*** 0.03** 0.07*** -0.01 Rice 0.14*** 0.09*** 0.06*** 0.14*** 0.09*** 0.11*** 0.12*** -2.24*** 0.12*** 0.12*** 0.09*** 0.12*** 0.05** -0.05** Noodles -0.05** 0.00 0.02 0.06*** 0.00 0.02 0.06*** 0.11*** -1.27*** 0.00 -0.02 0.10*** -0.05*** 0.05** Tomatoes 0.05*** -0.02** -0.01 0.07*** 0.02** -0.01 -0.01*** 0.04*** 0.00 -1.19*** 0.00 0.00 0.03*** 0.03** OrangeV 0.03 0.04*** -0.03* 0.11*** 0.02*** 0.01 0.08*** 0.09*** -0.02 0.00 -1.33*** 0.04*** 0.04*** -0.07*** Potatoes 0.05*** 0.03*** 0.02** 0.07*** 0.01** 0.03*** 0.02*** 0.05*** 0.04*** 0.00 0.01** -1.41*** 0.02*** 0.04*** OtherV -0.01 0.02** 0.04*** 0.07*** -0.02** -0.02*** 0.07** 0.03** -0.04*** 0.04** 0.03*** 0.03*** -1.32*** 0.08*** SaladV -0.15* 0.03 -0.01 0.53*** 0.12** 0.34*** -0.34*** 0.56*** 0.02 -0.24** -0.03 0.44*** 0.05 -2.23***

Note: Level of Statistical Significance - *** = 1%, ** = 5%, * = 10%

23

23

Table 6: Total Meat Expenditure Elasticities: High Income Households

Variable Total

Expenditure Beef 1.02*** Pork 1.03*** Fish 0.98*** Poultry 1.02*** Processed 0.97*** Cheese 0.92*** Bread 0.93*** Rice 1.03*** Noodles 0.99*** Tomatoes 0.99*** OrangeV 0.99*** Potatoes 1.03*** OtherV 1.01*** SaladV 0.92*** Note: Level of Statistical Significance - *** = 1%, ** = 5%, * = 10%

24

24

Table 7: Compensated Price Elasticities: High Income Households Variable Beef Pork Fish Poultry Processed Cheese Bread Rice Noodles Tomatoes OrangeV Potatoes OtherV SaladV Beef -1.16*** 0.07*** 0.08*** 0.14*** 0.03*** 0.04*** 0.04*** 0.11*** 0.02*** 0.11*** 0.08*** 0.15*** 0.03*** 0.26*** Pork 0.04** -1.20*** 0.04*** 0.05** 0.09*** 0.08*** 0.16*** 0.13*** 0.05*** 0.01 0.10*** 0.18*** 0.07*** 0.21*** Fish 0.13*** 0.08*** -1.23*** 0.14*** 0.08*** 0.04*** 0.05** 0.11*** 0.07*** 0.02 0.04** 0.15*** 0.09*** 0.23*** Poultry 0.09*** 0.06*** 0.06*** -1.37*** 0.07*** 0.05*** 0.12*** 0.11*** 0.07*** 0.13*** 0.11*** 0.17*** 0.09*** 0.25*** Processed 0.01 0.16*** 0.10*** 0.18*** -1.35*** 0.00 0.10*** 0.18*** 0.04** 0.11*** 0.10*** 0.15*** 0.01 0.22*** Cheese 0.06** 0.16*** 0.05** 0.15*** 0.01* -1.54*** 0.17*** 0.21*** 0.08*** 0.02 0.08*** 0.20*** -0.01* 0.37*** Bread 0.04* 0.18*** 0.05*** 0.23*** 0.07*** 0.09*** -1.47*** 0.13*** 0.09*** 0.04 0.12*** 0.13*** 0.11*** 0.19*** Rice 0.20*** 0.17*** 0.11*** 0.24*** 0.14*** 0.15*** 0.17*** -2.17*** 0.17*** 0.17*** 0.16*** 0.23*** 0.09*** 0.18*** Noodles 0.00 0.08*** 0.06*** 0.15*** 0.04*** 0.05*** 0.11*** 0.16*** -1.22*** 0.05* 0.05*** 0.20*** -0.01 0.27*** Tomatoes 0.11*** 0.06*** 0.03*** 0.17*** 0.06*** 0.03*** 0.04*** 0.10*** 0.05*** -1.13*** 0.06*** 0.10*** 0.07*** 0.25*** OrangeV 0.09*** 0.12*** 0.01** 0.21*** 0.07*** 0.05*** 0.13*** 0.15*** 0.03 0.05** -1.26*** 0.14*** 0.08*** 0.15*** Potatoes 0.10*** 0.12*** 0.06*** 0.17*** 0.06*** 0.06*** 0.07*** 0.11*** 0.09*** 0.05*** 0.08*** -1.30*** 0.06*** 0.27*** OtherV 0.04*** 0.10*** 0.08*** 0.17*** 0.03*** 0.02** 0.12*** 0.09*** 0.01 0.09*** 0.09*** 0.14*** -1.27*** 0.30*** SaladV -0.10 0.11* 0.03 0.62*** 0.16*** 0.37*** -0.30*** 0.61*** 0.06 -0.19* 0.03 0.54*** 0.08 -2.03***

Note: Level of Statistical Significance - *** = 1%, ** = 5%, * = 10%

25

25

Table 8: Uncompensated Price Elasticites: Low Income Households

Variable Beef Pork Fish Poultry Processed Cheese Bread Rice Noodles Tomatoes OrangeV Potatoes OtherV SaladV Beef -1.56*** 0.04*** 0.06*** 0.10*** 0.00 0.02** 0.07*** 0.09*** 0.02* 0.03 0.03* 0.07*** -0.01 -0.01 Pork 0.04*** -1.53*** 0.00 0.01 0.04*** 0.01 0.02 0.08*** 0.02* -0.01 0.05*** 0.12*** 0.04*** 0.04*** Fish 0.11*** 0.01 -1.32*** 0.02 0.09*** 0.00 0.07*** 0.04** -0.01 0.00 -0.03* 0.07*** 0.05*** -0.10*** Poultry 0.06*** 0.01 0.00 -1.61*** 0.05*** 0.02*** 0.05*** 0.09*** 0.02** 0.06*** 0.04*** 0.07*** 0.03*** 0.05*** Processed 0.00 0.09*** 0.10*** 0.15*** -1.74*** -0.04* 0.04** 0.08*** 0.00 0.07*** 0.08*** 0.13*** 0.05*** 0.03 Cheese 0.08*** 0.05*** 0.02 0.11*** -0.04 -1.53*** -0.02 0.08*** 0.03 0.12*** 0.07*** 0.09*** 0.02 0.11*** Bread 0.11*** 0.05*** 0.06*** 0.14*** 0.04*** -0.01 -1.48*** 0.05*** 0.02 0.11*** 0.11*** 0.07*** 0.05*** -0.02 Rice 0.10*** 0.08*** 0.01 0.15*** 0.04*** 0.03*** 0.03 -2.43*** 0.06*** 0.12*** 0.02 0.17*** 0.02 0.28*** Noodles 0.05** 0.05** -0.01 0.07*** 0.00 0.02 0.02 0.09*** -1.40*** 0.01 0.06*** 0.10*** 0.01 0.04** Tomatoes 0.03 -0.01 0.00 0.09*** 0.04** 0.05*** 0.08*** 0.11*** 0.00 -1.58*** 0.02 0.10*** 0.00 0.02 OrangeV 0.04** 0.06*** -0.03** 0.08*** 0.05*** 0.03** 0.09*** 0.03** 0.04** 0.03 -1.58*** 0.08*** 0.03** 0.01 Potatoes 0.04*** 0.08*** 0.02*** 0.07*** 0.04*** 0.02*** 0.01 0.10*** 0.03*** 0.06*** 0.04*** -1.56*** 0.02*** -0.01 OtherV -0.01 0.06*** 0.04*** 0.08*** 0.04*** 0.01 0.04** 0.04*** 0.00 0.01 0.04*** 0.07*** -1.39*** 0.01 SaladV 0.02 0.05*** -0.09*** 0.13*** 0.07*** 0.10*** 0.00 0.07*** 0.04*** 0.03 0.01 0.10*** 0.05*** -1.47***

Note: Level of Statistical Significance - *** = 1%, ** = 5%, * = 10%

26

26

Table 9: Total Meat Expenditure Elasticities: Low Income Households

Variable Total

Expenditure Beef 1.04*** Pork 1.06*** Fish 1.00*** Poultry 1.07*** Processed 0.96*** Cheese 0.81*** Bread 0.69*** Rice 1.32*** Noodles 0.90*** Tomatoes 1.07*** OrangeV 1.03*** Potatoes 1.05*** OtherV 0.96*** SaladV 0.88*** Note: Level of Statistical Significance - *** = 1%, ** = 5%, * = 10%

27

27

Table 10: Compensated Price Elasticities: Low Income Households

Variable Beef Pork Fish Poultry Processed Cheese Bread Rice Noodles Tomatoes OrangeV Potatoes OtherV SaladV Beef -1.47*** 0.12*** 0.11*** 0.23*** 0.04*** 0.07*** 0.14*** 0.12*** 0.07*** 0.10*** 0.10*** 0.21*** 0.05*** 0.11*** Pork 0.13*** -1.45*** 0.05*** 0.14*** 0.09*** 0.06*** 0.10*** 0.10*** 0.07*** 0.06*** 0.12*** 0.27*** 0.10*** 0.15*** Fish 0.20*** 0.09*** -1.28*** 0.14*** 0.14*** 0.04*** 0.14*** 0.06*** 0.03* 0.07*** 0.03 0.21*** 0.11*** 0.01 Poultry 0.16*** 0.09*** 0.05*** -1.48*** 0.10*** 0.06*** 0.13*** 0.11*** 0.07*** 0.13*** 0.11*** 0.22*** 0.10*** 0.16*** Processed 0.08*** 0.16*** 0.15*** 0.26*** -1.70*** 0.00 0.11*** 0.10*** 0.04** 0.14*** 0.14*** 0.26*** 0.11*** 0.14*** Cheese 0.15*** 0.11*** 0.05*** 0.21 0.00 -1.50*** 0.04 0.10*** 0.07*** 0.18*** 0.12*** 0.21*** 0.07*** 0.20*** Bread 0.17*** 0.11*** 0.10*** 0.22*** 0.07*** 0.02 -1.43*** 0.07*** 0.05*** 0.16*** 0.15*** 0.17*** 0.09*** 0.06*** Rice 0.21*** 0.18*** 0.07*** 0.31*** 0.11*** 0.09*** 0.12*** -2.40*** 0.12*** 0.21*** 0.10*** 0.35*** 0.10*** 0.43*** Noodles 0.13*** 0.12*** 0.04** 0.18*** 0.04*** 0.06*** 0.08*** 0.11*** -1.36*** 0.07*** 0.11*** 0.22*** 0.06*** 0.14*** Tomatoes 0.12*** 0.07*** 0.05*** 0.22*** 0.09*** 0.09*** 0.15*** 0.13*** 0.05*** -1.51*** 0.09*** 0.25*** 0.06*** 0.14*** OrangeV 0.13*** 0.14*** 0.02 0.20*** 0.10*** 0.07*** 0.17*** 0.06*** 0.08*** 0.10*** -1.51*** 0.23*** 0.09*** 0.12*** Potatoes 0.13*** 0.16*** 0.07*** 0.20*** 0.09*** 0.06*** 0.09*** 0.12*** 0.08*** 0.13*** 0.11*** -1.42*** 0.09*** 0.11*** OtherV 0.07*** 0.14*** 0.09*** 0.20*** 0.08*** 0.05*** 0.11*** 0.06*** 0.04*** 0.07*** 0.10*** 0.20*** -1.33*** 0.12*** SaladV 0.10*** 0.12*** -0.04*** 0.24*** 0.11*** 0.13*** 0.07*** 0.09*** 0.08*** 0.09*** 0.06*** 0.22*** 0.10*** -1.37***

Note: Level of Statistical Significance - *** = 1%, ** = 5%, * = 10%